Coin Brief ENDE

Two draft BIPs would let a wallet host check that a hardware signer used the host's randomness

The lead item in Bitcoin Optech Newsletter #431, published on 13 November, is a proposal to standardise exfiltration-resistant signing for hardware devices. Connor Avila posted two draft BIPs to the Bitcoin-Dev mailing list.

The problem is old. A signing device chooses the nonce for each signature, so a compromised device could encode parts of its seed in those nonces and leak it through ordinary transactions. Several manufacturers already let the host contribute randomness to nonce generation, but the technique has never been standardised. Optech lists earlier attempts: Stepan Snigirev's proposal in 2020, Pieter Wuille's overview of the techniques, Moonsettler's protocol in 2024 and the recently merged BIP461, which identifies possible attempts by a device to leak key material.

Avila's first draft defines a two-round protocol for schnorr and ECDSA signatures in which the host supplies its own randomness and then checks that the device actually used it. The second draft adds three per-input PSBT fields so that wallets can run the protocol through the usual signing flow. The proposal comes with test vectors, a Python reference implementation and a C module for libsecp256k1. Sjors Provoost had already published an experimental draft for single-signer taproot key-path signing and, according to Avila, plans to align his implementation. Avila asked for review, noting that the cryptography has not been audited.

The newsletter also reports Anthony Towns's plan to mine the genesis block of testnet5, the proposed replacement for testnet4 specified in BIP95. Mining starts once mainnet reaches its difficulty adjustment at block 975,744, and the genesis block will commit to a mainnet block hash at or after that height so that nobody can mine in advance. Because testnet5's minimum difficulty is about a million times higher than testnet4's, the first blocks need ASICs, and Towns described a public pool anyone can point hashrate at. Node software will still need testnet5 parameters, BIP54 consensus cleanup logic, DNS seeds, faucets and explorers before the network is usable.

A third item explains why compact block relay performed badly from mid-July to mid-October, when 0xB10C's nodes reconstructed only 1% to 5% of blocks without requesting transactions. Nearly all missing transactions belonged to chains of rune mints, each about 100 vbytes paying about 10 sat, sent directly to some nodes without announcement and partly replaced by transactions spending P2A outputs from the middle of a chain. Mempools diverged from what pools mined.

Two draft BIPs would let a wallet host check that a hardware signer used the host's randomness
Two draft BIPs would let a wallet host check that a hardware signer used the host's randomness — Coin Brief

Why it matters

Anti-exfiltration addresses the one component a user cannot inspect. If the PSBT fields are adopted, a wallet could refuse to broadcast a signature that does not show the host's contribution, turning a trust assumption about firmware into a check performed on every spend.

Written by Victoria Shinder.